Grice on 'imperative' as an operator

Imperatives: linguistics vs. philosophy
MERIN, ARTHUR
Linguistics - An Interdisciplinary Journal of the Language Sciences , Volume 29 (4)
de Gruyter – Jan 1, 1991
Publisher Walter de Gruyter Copyright Copyright © 2009 Walter de Gruyter ISSN 0024-3949 D.O.I. 10.1515/ling.1991.29.4.669


"This article examines literature pertinent to informed study of imperatives."

"It indicates the main problems and the challenge they, in turn, pose to comprehensive semantices of natural languages."

"It also hints at conceivable alternatives suggested by empirical questions arising from the texts and aims to show how method matters."

"Section 1 outlines Hamblin (1987)."

Section 2 goes into detail on defining imperatives,

section 3 on their pragmatic typology, and

section 4 on relations between addressee and referent of grammatical subject.

Section 5 examines Hamblin's analyses of negation and coordination.

Section 6 outlines Kamp (1979) on or-coordinated permissions, notes difficulties, and raises yet more serious problems posed by and coordination.

Section 7 summarizes four semantically important aspects of imperatives.

For readers with a mathematical background it also sketches one conceivable way of turning observations and critique into something more constructive.

Throughout, every opportunity is seized upon to argue that (formal) linguistics and philosophy might profitably go their separate ways.

Enjoyable and fertile as their relations ma^ have been, linguistics and philosophy are uneasy bedfellows.

Nowhere more apparently so than over the matter of imperatives.

Ask a philosopher, and likely as not you will find out that imperatives are directives or, more traditionally, commands.

Try the linguist, and you hear of a syntactic sentence type or, more traditionally, inflectional morphology.

Linguists' imperatives, that is, expressions such as (1)

Talk!

are not covered in recent textbooks on formal syntax or semantics. How Linguistics 29 (1991), 669-702 0024-3949/91/0029-0669 $2.00 © Walter de Gruyter 670 A. M er in they relate to the vast philosophical literature on directives, rife with expressions such as

(2)

One ought to talk

or with kinematic, logico-linguistic analyses of sentences such as

(3) You may talk!

(Lewis 1979; Kamp 1979), remains an open question (Gazdar 1981).

Some of their formal properties are addressed by Hausser (1980), Schmerling (1982), and, most fully worked out, Huntley (1984).

Yet the current monograph1 on English imperatives (Davies 1986) is very consciously informal.

Why BE formal?

Theoretical linguistics contains methodological multitudes, and three answers come to mind:

(1)

because this is how bona fide sciences have come into their own; that is, in bidding farewell to the hindsight of common sense;

(2)

because computer intelligibility spells job security;

(3)

because, at the very least, the constraints of mathematical specification may help one ride out some of the vagaries besetting the scholar's brief life.

---

Charles Hamblin's posthumous book (Hamblin 1987) aims to clarify imperatives' semantic properties from a philosopher's perspective while taking account of their linguistics.

It falls short of success, but in exemplary fashion.

Problems of method are instructively related to difficulties with replicability of key data.

And not trivially so. H's grasp of nuance is often superb, and his is also a sterling record in formal semantics and pragmatics.

Before his death in 1985 H had been noted, among many other things, for thoughtful contributions to the axiomatic explication of nonmetric plausibility (1959) and to the history of argumentation (1970).

He was the first to offer a formal interpretation of argumentative speech acts in terms of changes to individual commitment states (1971) and a denotational semantics of questions in Montague grammar (1973).

The book under discussion (Hamblin 1987) all but abandons that formally minded past.

One can only speculate what prompted the author's shift of stance.

Lack of cross-references, subsections inaccessible by table of contents, parentheses for footnotes and patchy coverage of the literature2 suggest a script unready for publication.

Well over 80 typographical or clerical errors, erratic indexing, and a blurb which informs one that previous authors tried to reduce imperatives to questions, bespeak utmost economy of editorial care.

Print and paper (T. J. Press, Padstow) are uncommonly good, though.

Nuel Belnap's elegant tribute to the author emphasizes valuable philo- Imperatives Outline review of Imperatives This goes into some detail on chapters 3-7; parts of chapters 1 and 2 are discussed extensively later on. Chapter \,

'The varieties of the imperative' (1987: 1-45), opts for a liberally functional view of its subject.

One should not, says H, 'pussyfoot the purely grammatical distinction between imperative constructions and others that do the same job';

for example,

(4)

Nothing in excess

(5)

Passengers must not speak to the driver

(6)

Jones, I want you to take charge of 4 Division.

Only 'when an imperative is very indirect', for example,

(7) Johnny, the door is open!

672 A. Merin should one diagnose something other, here an 'imperative implication' (1987: 6).

---- or as I prefer, "You've got mail!"

----

H thus appears to ignore distinctions between morphosyntactic mood on the one hand, say imperative or indicative

(8) a.

Everyone walk!

b.

Everyone walks!, and typically associated pragma-semantic modality on the other -- say,

JUSSIVE

(a generalization of directive)

or declarative, if we adopt Lyons's (1977) terminology.

One might invoke philosopher's privilege, inspired by H's counsel that 'too nice a distinction between "sentences, utterances, locutions, speech-acts and acts of linguistic communication" is for the pedants' (1987: 219).

But the book is not merely a study of directives or jussives, and the labor-saving credo begets diseconomy.

Most of the examples are English imperatives in the grammarian's sense.

Native-speaker intuitions -- that is, experimental data -- engaging their peculiar properties are used as evidence in argument.

Yet so are incompatible intuitions that are proper to indicatives having (among others) broadly directive uses.

Argumentative coherence is thereby compromised.

Apparent attempts to restore it affect the raw data: H's linguistic intuitions are consistently idiolectal at crucial points.

Accordingly, in what follows I use 'imperative' for the grammatical category,

and 'Imperative' to report H's ambiguous usage. Judging by context and credo it

implicates applicability of his claims to imperatives.

On lower taxonomic levels, H draws distinctions by type of authority, accountability, or interest among (functionally defined) 'imperatives proper', set off, in turn, from similarly divided 'permissives' and 'paraimperatives'.


The latter is a catch-all category and includes promises, plans, intentions, etc.

Chapter 2, 'Grammar-logic' (1987: 46-96), builds on the assumption that it is 'artificial to separate grammar from elementary logic' and consists of short sections on (a) subjectlessness and implicit subjects, (b) explicit subjects, (c) verbal aspect, (d) let 's imperatives, (e) indirect speech, (f) negation, (g) conjunction,

(h) disjunction, (i) imperative questions, (j) time and possibility, (k) conditionals, (1) imperative inference, (m) reasons, and (n) imperative's causal nature.

The treatment of negation and coordination starkly illustrates the danger of assuming their linguistic semantics to be tractable in a nonmathematical setting (see section 5). In (c), H notes gradient unease and mandatory agentive interpretation of statives (be tall) or achievements (intrans. break) (1987: 54-57). But there is no reference to Dowty (1979). Accordingly, the relations of (c) to (f) and to the important distinction Imperatives between subject referent and addressee (chapter 4) remain a tenuous matter of feeling. I shall take up H's observations on these and other peripheral matters only where Davies does not cover the ground as well or better. Sections (i) -- avowed inessential to the argument -- (k), and (n) are less substantial and will not be dwelt on. In (j) H nicely observes that the now of (9) Do it now refers to some near future time, not to the moment of utterance as in indicatives (1987: 81). And in (m) he offers insightful remarks on parallels between support for Imperatives and for inductive predictions (1987: 9If.). They are high points of the book (see section 7) but remain sadly unutilized and are inaccessible by index and table of contents.

Chapter 3, Three reductionist theories' (1987: 97-136), casually rejects attempts to reduce statements to Imperatives (a proposal here associated with Russell) and at greater length criticizes reductions of imperatives to 'deep-structure' indicatives of the You will..., You should..., and / IMP you to ... variety; that is, to predictions, statements of deontic obligation, and performatives, respectively. Generative linguistics recognized the first two as untenable by around 1970.

For clearly reasoned destruction of the third, see Gazdar (1979: ch. 2).3 H proposes four tests of reducibility: practical interchange in discourse, equivalence in reported speech, pragmatic role, and, most importantly, logic.

His overall argument (as distinct from his conclusions) can only be as correct as the claims of chapter 2, since the fourth test is deemed crucial (1987: 99).

In particular, it is invoked to reject more abstract

'Imperative operator'

accounts which do not posit abstract verbs, for example, that of Hofstadter and McKinsey (H&M) (1939), who are classed as implicit You will theorists.

In relating their work H says, ' "A4" is written for Let it be the case that A' (1987: 101) -- a significantly inverted description of an axiomatically defined operator that represents an idealized (social) act, which H&M label fiat. Given his background, it is surprising that H should ignore the difference between paraphrastic shorthand for an ill-defined intuitive notion and intuitive gloss of an axiomatically defined and/or model-theoretically interpreted expression. What alone counts for the latter two are inferential relations and the constraints they place on concrete interpretations.

(See section 3 for H. A. Simon's model-theoretic development.) Chapter 4, 'Action-state-semantics' (1987: 137-166),

adopts a 'possibleworlds' semantics for propositional content, sorted into states and state changes (ordered pairs of states), the latter again partitioned into acts 674 A. Merin causally attributable to agents ('deeds') and mere happenings. All of these are ultimately defined as subsets of possible worlds (1987: 143). To distinguish proper -- that is, intended viz. 'intensional'

-- satisfaction of an Imperative from merely 'extensional', possibly accidental satisfaction,4

H introduces the notion of a strategy: a function allocating a deed to each time/context pair. Weakenings to 'partial strategies' (identified with disjunctions of complete strategies; 1987: 157) and relativization to imperfect information about the current context follow naturally. Notable is a threefold classification of deeds into those that (i) further a strategy, (ii) ensure dissatisfaction of the corresponding Imperative, and (iii) do neither. If deeds available do not present a choice of either, the addressee may follow a 'null strategy', that is,

do anything at all.

That predicament neatly serves to interpret the intuitive concept of an Imperative's having 'lapsed'.

The interpretation of an Imperative -- its 'addressee-action-reduction' -- is given by the weakest partial strategy (the disjunction of all partial strategies) which keeps the addressee within the set of possible worlds designated by the propositional content.

Floyd-Hoare semantics for program specification (Pratt 1977) is noted as an inspiration.

But there is no mention here of related model-theoretic studies (Lewis, Kamp, Simon); nor anywhere of Äqvist [Ä] (1974), a rigorous model-theoretic 'game-tree' analysis of action and causation and arguably the benchmark for any logical treatment of doing and forbearing.

A builds on Chellas (1969,1971), the standard account of Imperative logic, which H mentions as insufficiently diversified (1987: 138).

Like H, Ä does not extend his strategy model by the usual decisiontheoretic means for guiding or predicting choice of strategies: payoff functions reflecting ceteris paribus preferences viz. opportunity costs. Unlike H, Ä does not set out to explain phenomena whose major taxonomic coordinations they appear to constitute (see section 3). Chapters, The consistency of a set of imperatives' (1987: 167-199), examines commands clashing with commands ('quandary'), with permissions "contradicting' them, and with permissions Obstructed' by them. Discussion proceeds by reference to topographic relations on a 'deontic square' of oppositions with imperative labels (Do D, Don do D, Do D if you like, Don't do D if you don't like), rather than by axiomatic or standard model-theoretic means appropriate to invocation of possible worlds in chapter 4. The problem of how permissions change a set of prior deontic impositions is, in part correctly, identified as extending to 'withdrawal of locutions of other kinds' (1987: 71); but not addressed.5 H notes that permissions may be individually unobstructed though jointly obstructed Imperatives (1987: 190f.)· Yet, like much of the literature on dialogue and argumentation that took its cue from him, H does not address the ensuing logical problems which cited works (Cornides 1969; Lewis 1979) try to face up to; nor their conceivable linguistic import. Chapter 6, 'Imperfect ability' (1987: 200-217), argues that the 'paraimperative' correlate of 'knowledge that' is 'knowledge to' -- that is, 'ability' having an Owner' who is, to most intents and purposes, the analogue of the Imperative's intended agent.

H observes linguistic constraints shared by imperatives and to infinitives designating abilities. For example subjects are generally owners, modal auxiliaries are prohibited, certain adverbs and nonagentive verbs are poorly admissible, if at all, etc. He notes absence of 'permissive' analogues of abilities and suspects that this is because they already resemble permissives: they can be exercised (1987: 204).

For H, inspired by Vendler (1972), linguistic correlates of intentions are English to infinitives and that-someone-[do X] clauses, with parallels to undertakings and directives, respectively; These observations are original and an empirically useful supplement to Huntley (1984); see section 2.

*********************************** ON GRICE **************

Chapter 7, 'Imperatives in dialogue' (1987: 218-241),

attacks the Gricean program of taking mentalistic

notions such as beliefs and desires [boulomaic] and intentions as primitives in the philosophy of language and discourse.

Hamblin proposes, instead, that public commitments be fundamentals and that actual mental contents make no difference to what one says.6


He also diverges sharply from Wittgenstein's popular legacy of simple language games ('Slab!'). Rightly, I think, he notes that they and computational models of the kind immortalized in Winograd (1972) replicate an idealized master-slave or animal-command relationship in which accountability, requests, and permissives have no place (1987: 222).

Yet he does not observe that a model of nonaccountable master and faithful slave explicitly underlies Lewis's analysis of permissions. Neither does he ask, then, whether this might have something to do with the logicolinguistic problems attending it and similarly committed proposals. By means of an informal 'analytic shorthand' H sets about specifying the proper conduct of dialogues involving commands, refusals, acts of noncommitment, permissions, and conditional commands. The notation imposes no inferential constraints but neatly sidesteps what would have to be some counterintuitive dialogue fragments. Thus, justification of imperatives, demanded by Why? challenges, is stipulated rather oddly to be another imperative (1987: 234). There is no mention of coordinate imperatives such as

(10)

Give me your wallet or have your head blown off, 676 A. M er in which might indeed come close to fitting the typical bill on occasion. The implicit denial of a role for declaratives is also at odds with doubts voiced earlier on about Hare's contention (1952: 28) that no imperative conclusion can be validly drawn from a set of premises which does not contain at least one imperative. Hamblin argues that (11) It's lovely in the water, so come in does not convey a hidden Imperative such as (12) Always take advantage of good swimming conditions, on the grounds that inserting it 'would turn the whole exercise into a piece of bad grammar' (1987: 90). Even disregarding a likely confusion of syntax, denotation, and circumstantial inference, (11) is not shown to be an instance of valid inference on any notion of validity acceptable to a logician during office hours. A reconstruction relativized to a theory, stated in purely declarative terms and including background assumptions such as conveyed in (12), might well be. But this would imply a theoretical stance (see section 3) that H is committed to deny. Such problems apart, the rules do not engage, let alone predict, what should worry at least the Fregean linguist: a potential for unobvious influence of act type on apparent propositional content (Kamp, Lewis). They might, however, inspire the business engineer to formulate plausible protocols of communication; that is, abstracting from unobvious socioeconomic and linguistic fine structure, or taking for granted the availability of suitable canonical translations into formal languages.

2. Defining imperative What could justify the conflation of form and function?

(Disregarding for now the argumentative problems arising from H's ambiguous usage.) One might argue, as H rightly does, that inflectional morphology alone wj\\ not distinguish, say, English imperatives from other, semi-theoretically distinct nonfinite clause types. And then conclude with him that one cannot avoid appeal to context of utterance (1987: 46), that is, to apparent function. But one can distinguish; accessibly so for English imperatives. In brief (Davies 1986: 7if.), they lack tense inflection;7 their subjects are optional and, if present, pragmatically restricted; negation bearing on the main verb is don't or not preceded by do, which, in turn, may precede be or auxiliary have for emphasis.8 Purely syntactic definition thus appears feasible, independent of proto- Imperatives typical functional criteria used heuristically prior to definition. And it affords, in principle, a basis for semantic analysis in the broadest sense. But how to define semantics? One fairly austere tenet, reflecting the empirical linguist's objectives and mathematical usage might be this: specify a mapping from a domain of (better less than more) abstract syntactic objects to a domain of (more or less) abstract objects with enough relational structure of their own to account for robust, immediate-response intuitions on acceptability and paraphrase that are pretheoretically felt to be related to meaning. More forthrightly put: specify the kind of constraints on intuitions which your sense of aesthetics would not wish to build into the bare formation rules of a formal language. Such hard-core psychologism might benefit descriptive adequacy; but it is far from being common ground among students of meaning. H, in line with much pre- and post-Montagovian philosophy of language, short-circuits to the elusive reality of what is really meant. There is, to be sure, little nontrivial consensus on what, if anything, distinguishes the meaning of (grammarians') imperatives from that of expressions in other moods. Perhaps under the spell of'illocutionary force', the question is often conflated with another: what distinguishes various act types commonly performed by means of imperatives. Thus, for H (who treats mood now as a semantic notion) ... it is also clear that divisions by mood and force mingle arid overlap. SUGGESTIONS and ADVICE, for example, can be neither indicative or imperative, and in some cases they seem to straddle the line, in the same way as moral or legal utterances. The same applies to WARNINGS, which can be imperative [Don't go down to the woods today] or indicative [// could be stormy later] (1987: 23).


Clearer, neo-Fregean discussions of force, mood, and forceless content are in Ross (1968); and in Hare (1970), to whom is due the terminology of neustic (speech-act viz. force-indicating operator), tropic (semantic mood-indicating operator), and phrastic (force- and moodless propositional content) adopted by Lyons (1977).

---- also "clistic".

---- "Some sub-atomic particles of logic" -- Hare cites Grice extensively in 1967 _Mind_.

---



Hamblin apparently mistakes 'tropic' and 'neustic' as alternate labels for the same thing (1987: 102).

Yet his accompanying remark on dearth of logical detail in Hare's proposal makes a telling point.

A distinction without a formally specifiable difference is merely asking to be overlooked.

Hausser (1980) represents both a formal advance and continuity with the linguistic tradition.


It is, he writes, 'a mistake to implement speech act properties in the semantic characterization of syntactic mood'.

The denotatum/designatum of an imperative, say

Leave!

is identified as 'a property (roughly that property which the speaker wants the hearer to acquire)'. 678 A. Merin

In Montagovian PTQ formalism it translates ~[2{} & leave'()], where 2 indexically designates the property of being the hearer.

This engages three usefully separable issues:

(a) syntactic analysis and attendant 'internal' semantic choices required to satisfy compositional type compatibility;

(b) 'external' semantic properties of whatever object (here: proposition) Hausser's property would abstract from; (c) pragma-semantic problems shelved by use of the adverb 'roughly'.

But, as Huntley (1984) notes, Hausser's proposal will not do for imperatives with syntactic subjects (especially third person; see section 3), nor relate them to inflectionally cognate constructions.

His own modeltheoretic attempt to relate imperatives and embedded to or that infinitives proceeds differently on both of (a) and (b).

Indicatives denote propositions indexically linked to the actual world of evaluation (no matter whether past, present, or future) and typically embed under epistemic/ alethetic verbs.


The nonfinite forms denote propositions that are still free to characterize another future or possible world and typically embed under deontic/boulomaic verbs.9



As regards issue (a), there is, I think, little point in discussing whether apparently subjectless imperatives have subsurface 'syntactic' subjects without commitment to BOTH a formal syntax and a compositional semantics.


Neither Hamblin, nor Davies, nor Beukema and Coopmans are so committed.

Huntley treats imperatives as sentences denoting propositions and, in consequence, is committed to syntactic-deletion transformations. Not an attractive solution for categorial or generalized phrase-structure grammarians.

They would note that, once an interpretation for actual or hypothetical sentential forms has been settled upon, the next move is to try and interpret phrases by functional abstraction. Must a nondeletionist opt for the base-generated VP analysis of the subjectless form, then? The simplest alternative I can think of is to extend languages' vocabulary by a large, finite set of autoiconic symbols whose cardinality at any time is as well known as that of the set of truly proper person names. You, rather than a deep you, would occupy the syntactic subject position of an apparently subjectless imperative mood string addressed to you. And you would denote yourself or, better, the maximal consistent set of properties uniquely associated with you. This proposal10 will surely sound dismaying to the ambitious lexicographer, parochial to the 'pro-drop' analyst, and reminiscent of the wilder shores of semiology. But it gets by, locally at least, without trading syntactic transformations for homophones or helpful spirits that insert, move, or delete hidden contextual variables such as 2. However that may be, syntax appears to determine semantic mood (viz. modality) for English, IF we restrict the notion to interest- and Imperatives incentive- or warrant-neutral aspects of meaning; more generally, to those which abstract from essentially politico-economic relations. On points (b) and (c), Huntley's work might suggest that 'tropic' should be taken even further apart: into (i) properly modal, fairly asocial components (actual vs. nonactual), and (ii) the social, though economically neutral, components implicit in Hausser's gloss of property acquisition.

Note that Hausser's wording still imports a 'neustic' component.

To be neutral with respect to claim/concession polarity, his 'wanf would require the marked, nonspecific reading, which must encompass the subgloss "be willing to\ Idle pedantry? Davies (1986: 51) argues that 'semantically an imperative sentence constitutes simply the presentation of a proposition, representing a potential state of affairs'. One might extract from this a distinction akin to Huntley's: potential is, after all, nonactual.

But her failure to distinguish sentence and speech act precludes benefit of the doubt.

While the speaker who utters a declarative which asserts a proposition p is conventionally assumed to accept that p is true, the speaker who utters an imperative which presents a proposition p is conventionally assumed to accept its being made true.


Imperatives get their directive force as indirect speech acts.

"Uusing an indication of one's acceptance of something's being done in order to communicate one's intention that it be done" (1986: 5If.).

This relies excessively on the explanatory magic of indirect speech acts.

'Acceptance' introduces a bias which will have to be neutralized before admitting respective distinctions between commands and permissions, between demands and concessions, or between assertions and admissions.

Neither will it distinguish imperatives' denotata from acts of suggestion, which might, in Davies's terms, lay equal claim to being mere presentations.

3.


Pragmatic typology of imperatives H subclassifies Imperatives' import in several ways but does not exploit the symmetries or potential for independence of his categories.

To show why and how, I shall stay close to his exposition rather than follow a more obvious systematic one.

There is a promising start: H notes

(i) type of authority carried by Imperative utterances, and

(ii)

aims and expectations putatively entertained by their issuers and recipients (1987: 5).

Contrast along the first dimension distinguishes commands from requests; both of which are 680 A. M er in issued 'in the putative interest of the utterer, independently of the interest of the addressee'(1987: 9). 'Putative' is important and what the linguist should, I believe, also stick to. After all -- conventional default, rather than philosophical generality admitting 'real' exceptions, might condition spontaneous intuitions. If taxonomy is not to remain idle, academic treatment of speech acts might engage it in pursuit of two potentially distinct objectives: the philosopher's as an abstraction from judicial discussion ('explication'), and the linguist's as an instance of psychological inquiry into a circumscribed class of data ('explanation').

Granted focus on the latter, a claim of independence from addressee interests is too strong. To see why, note that H maintains that 'essentially and typically the object of a command or request is to get the addressee to do or bring about something the issuer wants done, which the addressee would not be disposed to do otherwise' (1987: 9).

Sociologists will recognize the standard preliminary definition of a power situation, formally treated in the theory of bargaining games (Nash 1953).

But the linguist attuned to 'negative raising' may observe that not to be disposed or inclined to do something usually amounts to being disposed or inclined not to do it. Indeed, since Weber (1922: section 16), power has been routinely defined in terms of the difference which its exercise would make to the probability of an action's performance in the face of resistance. Surprisingly, then, and even though H associates demands with bargaining, he fails to consider a putative or nonnegligibly probable DISINCLINATION of the addressee as a potential structural feature of speech-act types. As might have been advisable to preserve coherence. With invitations 'the interest', he says, 'is mutual, and we shall have to put them in a different or borderline category' (1987: 10). Yet the taxonomy puts nothing to the other side of the borderline containing commands, demands, and requests. That region would, presumably, contain 'some parallel or fringe entities', exemplified, for example, by (13) All right, go the football match, (if you like/but don't expect me to be here when you get home), any of which 'withdraws or denies some bar or objection' (1987: 29). These are, according to H, permissions, concessions, or authorizations and are labeled 'permissives'.


Hamblin notes the long tradition of deontic 'squares of opposition' (1987: 31), patterned after the familiar quantificational and modal squares usually labeled 'Aristotelian', and the 'complementarity' (presumably H meant to write 'duality') of propositional operators Obligatory' (O) and Imperatives 'permitted' (P).

In near-parallel to what he calls 'imperatives proper', H discovers 'command-', 'request-', 'demand-', 'advice-', 'invitation-', and 'recipe-permissions'.

But no empirical justification for the last three is offered; nor consideration of a plausible alternative: that command, demand, and request intuitively correlate with permission, concession, and authorization, RESPECTIVELY.

Accordingly, H fails to consider -- as emphasis on putative social relations would counsel -- that in the Stereotypie case each of the latter three act types exhibits the INVERSE interest structure of its correlate in the former group. Which suggests a common distinction along the 'interest' dimension. Having eschewed symmetry in the plane of interests, H does likewise in that of warrant, authority, or incentives.

Commands, demands, and requests are dubbed 'wilful' (utterer-boulomaic)

Advice, instructions, and recipes are, by contrast, 'non-wilful' (or rather addressee-boulomaic).

Advice, instructions, and recipies are 'concerned not with the interests or wishes of the utterer, but with those of the addressee'. They carry 'rational' rather than issuer's 'social' authority.

In Kurt Baier's finely tuned spelling, their authority is 'principle-related' rather than 'principal-related' (1987: 10).

This is an important distinction, exploration of which must surely take one into deep questions of philosophy and sociology.

H stays closer to the plausible. His paraphrase, 'rational' vs. 'social', suggests that rationality criteria (as distinct from Platonic formal systems enshrining them) are not social.

Concomitantly, the allegation of addressee bias confuses the relatively and ostensibly DISINTERESTED nature of advice, etc. -- on which the transpersonal nature of argumentative support might be thought to hinge -- with that of permission, concession, and authorization.

What could have motivated the taxonomic bias?

Nonwillful Imperatives, abundant in applied science, share declaratives' rational ACCOUNTABILITY.

H concludes that they must have a logic.

Conflating Kant's 'rules of skill' with the defmitionally superordinate class of 'hypothetical imperatives' (1987: 16),11

he argues that technological imperatives are not reducible to cause-effect statements such as

(14)

If you do Y, you will achieve X,

since we cannot usually specify with sufficient accuracy.

This is of doubtful pertinence.

Consider the familiar ceteris paribus conditions underlying.

(15)

If this match is struck, it will light (unless you have wet it or not struck hard enough or ...)

Problems are no different in kind here, and manageable for anyone who 682

A. Merin can live with the 'frame problem' (McCarthy and Hayes 1969).

Indeed, problem similarity emerges clearly in Kamp, Stalnaker, and Lewis, with consequences for inferential intuitions on natural-language expressions.

H next asserts that nonwillful Imperatives' logic differs fundamentally from that in current use for declaratives (for evidence offered, see section 5).

This is tacit contradiction of Simon (S) (1965), who shows by worked examples how to move between positive descriptive theory and prescriptive task specification.

S introduces a 'command variable' bound by an 'imperative operator' ! which one might gloss 'set to satisfy ', where is free in .


Conversion of existential statements in pure theory, for example,

(16) There is a production quantity x, which equals x*, the quantity maximizing profit, into applied nonvacuous commands, for example, (17) Set production quantity equal to x*, the quantity maximizing profit, is possible provided (i) the statement is not a (definitional) identity in the command variable, x, and (ii) is under the addressee's control.

These conditions need not be expressly imposed for the other direction.

Inference proceeds wholly in the declarative mode, if necessary by intermediate reconversion.


Constraints are combined by the usual set-theoretic operations. S does not purport to address linguistic points of grammar or usage. H's brief extends beyond the impersonal aspects of optimizing conduct.

But the 'nonwillful' case, which motivates H's demand for a new logic (rather than for other kinds of comprehension routines), is essentially the one addressed by S. The claim, central to H's argument, might have benefited from discussion of its advance contradiction and from reflection on analytic objectives.

In the linguistic domain, H's premature conftauon of Vwo subtly related socioeconomic parameters -- (i) distribution of interest and (ii) type of warrant or, more generally, incentive for accepting a deontic or epistemic commitment -- just throws away potentially useful structure. It also obscures the nature of deviant cases. Wishes, such as (18) Get well!, are said to be akin to invitations, being in both parties' interest. H is rightly doubtful about (19) Go to hell Imperatives but might have got a positive lead by considering its heavier relative, (20) May the devil take you. Davies (1986: 57) notes that wishes are not (add: ostensibly) in the addressee's power to realize and grants them a status that Schmerling (1982: 212), staunchly upholding the Chomskyan doctrine of asocially structured language, assigns to the use of imperatives in general: relics of belief in mysterious, magical powers.

Another distinct category for H are 'formal' viz. 'rhetorical' imperatives such as (21) Let be the length of the fish (22) Consider what would happen if... (23) Suppose the letter has gone astray. Rather than consider to what extent ostensible disinterestedness might condition their linguistic or philosophical properties, H states apodictically and inconsequentially that there is no 'act' the addressee is being asked to perform, not even a mental one (1987: 25).

This illustrates again a point easily lost in more prestidigitatious hands: that one might usefully separate philosophical questions (such as, Do we need to speak of mental acts? Could we coherently do so?) from linguistic and psychological questions (such as, Do we get better predictions in assuming that putative acts are being conventionally enjoined of the addressee?). Quite late (1987: 26) H notes that many imperatives, for example,

(24) Bring in some fresh bread when you come home.

said to one's spouse, are unspecific between any of the main headings.

The linguist would take this as the starting point of analysis, not as a minor interim qualification. It is well known, at least since Hermann Paul (1885: section 93), that speech-act types assigned to imperatives encompass all those considered above, and more, and that they vary with context and content-based constraints on context.12 Quirk et al. (1985: 11.29) perceptively note that the two dimensions of interest and authority (one might amend authority to: authority or power) play a major role in contextual specification.

What the literature does not consider are possible implications of their ubiquity for languages' semantic structure. H finally introduces a third 'catch-all' category of 'para-imperatives', which contains

(i)

promises, vows, and undertakings (J. L. Austin's 'commissives');

(ii)

plans, projects, proposals, schemes, programs, and designs -- these are 'forceless';

(iii)

mental acts and states accompanying the proposed actions of (para-)imperatives; and

(iv)

persuasions and incitings (whose status remains unclear).

Let's imperatives calling for 684 A. Merin joint action are characterized as a complex blend of Imperative and undertaking (1987: 35).13 (H observes that undertakings are never imperative but does not ask why.) Plans only acquire force when someone orders, requests, advises, or invites putting them into effect. One might infer that 'strategy' is a synonym for 'plan'. 4. Scope of responsibility Davies (1986: ch. 5) specifies conditions on explicit subjects -- roughly, a contextual need to pick out the right (subset of) potential addressee(s) -- and on the effects of admissible flouting.

Thus, redundant "you" carries an implicature of speaker's authority.

Much like Davies, H uses the prosodically coded distinction between a vocative and an imperative subject14 (1987: 52), as in (25)

(26)

John, move upstage John move upstage, to distinguish logical subject (who is to carry out the action specified) and addressee (who is usually responsible for bringing about that it is carried out).

He proposes 'perhaps the most fundamental principle of the book',


'addressee-action-reduction' [AAR]:

the meaning of any clearly specified directive can be spelled out in 'plain-predicate' (overtly subjectless) agentives (1987: 58), that is, in terms of what actions it enjoins of the addressee(s).

This makes a very important point: that one should distinguish any ACTION(S) ENVISAGED of the logical subject from the ACT ENJOINED of the addressee.15

Hence, H's occasionally dubious insistence on cases that ought to reflect their independence: What a speaker intends to permit is often that a specific person be given discretion what is done.

Compare

[(11)}

Henry go to the party, if you wish

with [(28)]

Henry go to the party, if HE wishes



The first gives the discretion to the addressee, the second to Henry. The simpler [(29)] Henry may go to the party does not separate the alternatives, and is usable when it is not necessary to do so (1987: 31).

Now, neither (27) nor (28) are acceptable intersubjective English on Imperatives the typographically suggested (and intended) prosody which ensures that Henry is not the addressee.

What H ignores to the extent of generating a proper idiolect is that imperatives are distinguished from indicatives capable of assuming a coextensive pragmatic function by the ABSENCE of such ambiguity.16

Still, AAR has the edge on what is available over the counter elsewhere. Davies, for example, sees (intersubjectively acceptable) English utterances of the kind engaged by H as cases 'where the motivation of addressing an order to someone other than the intended agent seems to be to get the addressee to report the directive to the third persons concerned' (1986: 140f.). But her examples,

(30) (31) Your men guard the front while we keep round to the back These children of yours keep out of the garden or I'll set my dog on them, 686 A. Merin unrealistically strong or empty. None of the problems its nonvacuous interpretation might raise are discussed; nor its desirability as an engineering principle in the light of recent computer science's emphasis on 'barriers of abstraction'.17

Once the pragmatics of compliance are taken seriously, we must, indeed, (i) constrain the standard of performance of action envisaged18 and (ii) specify the addressee's concomitant responsibility to ensure that the means employed do not violate the speaker's interests more than is necessary for performance of (i).

H, in pursuit of the procedural red herring, does not consider that a 'how to' will comprise both points.

Neither does he note that, on both counts, computations of utility tradeoffs will be required to select from the usual indefinitely large variety of means employable. In short, what matters is WHAT is being brought about or performed ALL THINGS CONSIDERED.

To repeat, for the linguist all this would seem pertinent if, and only if, it can be shown to be required for the explanation of phenomena on home ground: native-speaker intuitions on acceptability and paraphrase.

To descriptive formal semantics this means, not least, intuitions which engage the 'closed class' recursive component of meaning. 5.

Connective semantics H claims that 'five different and independent negations' may apply to Imperatives (and, it seems, imperatives) (1987: 64), yielding 32 different affirmative/negative forms (1987: 71) for a given stative predicate S: (32) (No commitment as follows:) (Refrain, if you wish, from) (not) f (making) | it be the case that (not-)S. {(letting) J Type 1, such as don't applied to 'stative' predicates such as

(33) Be here at lunch!

specifies the complementary end-state. Type 2, applied to action-specifying predicates, changes an instruction (34) (35) Run to carry out that action into one to refrain from doing it Don't run. Imperatives Problems, according to H, arise when (as is usual) the 'stative' (33) is understood to mean roughly (36) (37) (38) Take steps to be here at lunch, Don't be here at lunch Don't take steps to be here at lunch. whereas does not mean (roughly) Yet, though H concedes that usually there is no ambiguity, he maintains that we sometimes wish to negate (33) in just this second way. It is not clear who would wish to do so, nor how the wish could be realized. The reader should confirm that at most a concessional (39) All right, don't be here ... is intuitable, presumably involving something like types 4 or 5.

Paraphrastic 'steps' are, I think, no substitute for best-practice decompositional analysis; point of departure for which is Dowty (1979). Type 3 is said to be periphraseable as in (40) Don't actually kill him. meaning that if he were to walk under a bus without your agency, that would be fine (1987: 66). But empirically it is not clear that don't -- with or without interposed actually -- will induce the desired meaning. H uses undefined predicates Make and Let to symbolize commission and omission and admits a third possibility: doing neither. More usefully, Äqvist (1974) identifies that option with absence of a decision point for the actor in his game-tree model.

Applied to linguistic ends, this predicament will preclude felicitous imperatival utterance. It would therefore predict that the putative reading, far from characterizing directives, distinguishes them from declarative reports by its absence.

Type 4 looks like the deontic logician's 'external' negation.


It is to turn a prohibition ('internal', type 2 negation of command) into a permission. This leads to a 'deontic square' with imperative mood captions (1987: 69). But H leaves unexplicated how negation could apply to act types in use. And when he avers that an invitation has 'both faces' of command and permission, one is left wondering what kind of deontic Necker cube it is supposed to live in. Type 5 indicates speaker's noncommitment. H notes that it is 'not especially characteristic of imperatives'. In fact, one would be hard put to elicit it as other than the negation of an explicit performative verb 688 A. Merin (preferably in the present continuous tense) of suitable indicative sentences. In which case it is not clear that it is not common-garden-variety main-verb-scope negation. Disjunction (or), according to H, has two senses that give 'relatively little trouble'.

This may be news to deontic logicians who have cut and sometimes lost their teeth on Ross's paradox (see Kamp 1973 for concise appreciation).

----

Standard deontic logics license the inferences from (41) It is { ^fTM1 C > that this letter be F posted v ' [obligatory] to (42) (41) or burnt.

In striking, attractive, (and dubious) imperatival paraphrase this leads from

(43)

Post this letter!

to

(44)

Post this letter or burn it!

The crucial distinction involves ambiguity between a 'choice-offering' (CO) or, granting choice among alternatives, and an 'alternative-presenting' (AP) or, leaving choice to be determined by further instructions.

H appears to paraphrase qvist (1965), who represents '

(45)

It is obligatory that or as plain

(46) ( V ) and the 'CO' version as (47) 0+(,) = df ( V ) & () & ().


If one's objective is to map fragments of a natural language to a logic in which certain counterintuitive inferences are not derivable, consistent ad hoc stipulation of homophony is all right. But not, I feel, for elegant description of nonparochial aspects of language. appears compelled into hard-to-replicate perception. He finds that

(48)

Take her to Knightsbridge or Bond Street will have both readings (1987: 73). I cannot hear '. Unlike, say, for (49) You may take her to Knightsbridge or Bond Street, Imperatives which readily exhibits, as Kamp (1973, 1979) recognizes, a reading that licenses continuations such as (50) but I haven't yet decided which.19 Continuations such as (51) but first consult my secretary on which one (52) but not the more expensive of the two are also admissible with (48). But one should argue -- and could, if the resources of a properly procedural, kinematic semantics were available -- that they now remove a freedom of choice provisionally granted, much like a postposed except clause.

Why?

Because the first continuation should be admissible, too, if alternatives were merely presented. And, unlike (49), (48) does not take an AP interpretation on its own. H adds that simple permissives, like (53) Smoke, if you wish, are interpretable similarly as '{You may smoke, You may refrain from smoking}' and observes that or, in permissions, commonly indicates choice among exclusive options, 'though conjunction is more nearly indicated than disjunction' (1987: 76). What binary operation, if any, this familiar stab in the dark points to (compare Ehrenkranz 1973) is what analysis of imperatives should attempt to clarify. It is one thing to observe local paraphraseability by and\ another to come up with a compositional semantics using Boolean conjunction which will not predict that only both conjuncts together are permitted (see section 6). H bypasses that problem but detects two conjunctions in imperatives (1987: 73). The 'separable', more basic, enjoins partial fulfilment as a second-best; the 'inseparable' doesn't.

He credits Menger (1939), who had noted that one might entertain a wish for a pair of complementary goods (say, a cigar and a match) that are no use to one singly without the other.

But he overlooks20 that Menger distinguished wishes (desires, /)) from commands (norms, C) by the very invalidity of D schemata that he held to be valid with C: (54) a. b. C(0&©) «-> (&&) -» C(0)&C(0) £>() &/)(©) (1934: III, 10 [1974: 36ff.]; 1939: 59).-Hence, unlike wishes, commands must license the inference to either of the right-hand-side conjuncts singly.

Menger also points out that we might desire and desire but not their joint realization.

And that this is a matter of world knowledge, not of logic (1934: III, 11). 690 A. M er in It is a large step from noncommittal desires to acts intended to commit others directly to action.

By typical intuitions, if

(55)

Do A and B!

is taken as a demand, request, or command, it is satisfied -- that is, complied with -- if, and only if, both A and B are.

H appears to sacrifice descriptive aims for those proper to applied ethics.

A similar distinction of possible interpretations is a problem for description of concessions or permissions (see section 6); but this is not part of the book's claims.

It is unfortunate that H should refer to his avowedly burgeoning conception as the logic' of imperatives.

The occasional shorthand notation is less insistent than ordinary language on making underlying commitments explicit.

There is no recursively specified syntax or semantics.

Indeed, doing 'logic by case-law' (1987: 85) appears to deny specifiability of a statutory consequence relation -- that is, of a logic. (As distinct from assorted inferential heuristics or other formalized theories having inferential uses on the side.)

But why object to a little abuse of language?

Because it obstructs possible realization that one may be in a wholly different, though well-structured, ballgame. 6.


Kamp's permission problem and its extensions Point of departure for rigorous analysis of coordinating connectives is Kamp (1979), who does not address the imperative mood but aims for an elegant explanation of sentences such as (49).

Committed to (i) syntactic and semantic compositionality and (ii) a uniform Boolean interpretation of or as set-theoretic union (disjunction),21 he considers several theories.

Any one of them, he argues, might be a candidate for informing a comprehension strategy under suitable circumstances. The theory which appears to satisfy more of his syntactic and pragmatic desiderata than others treats permission indicatives as having primarily a report reading.

Any directive use, as well as the appropriate interpretation of propositional content, would then be derived by an implicature yet to be accounted for. For imperatives, however, this line of credit is unavailable,22 and one would have to adopt Kamp's primarily directive analyses.

Permissions, according to Kamp, drawing on his earlier (1973) paper and Lewis (1979), expand the set of their ultimate addressee's admissible options for action.

A command will subtract from the set of option worlds by set intersection with its propositional content. Dually, a permission adds to the current set of option worlds (by set union) the set of Imperatives option worlds designated by the permission sentence's propositional content. It will thereby also add wholly unintended worlds to the option set. Thus, Ego's permission to Alter to drink Ego's wine. (56) You may take my wine, will also add worlds where Alter takes Ego's wife as well, etc., since some of those worlds will be contained in the set of (all those) worlds where Alter drinks Ego's wine.

To filter out unintended worlds, Kamp opts for an ordering scheme proposed to Lewis by Robert Stalnaker (n.d.) and analogous to a widely accepted one for conditional antecedents.

Options are completely preordered (reflexively, transitively, and connectedly) in 'concentric' nests of equivalence classes of ever more 'reprehensible' worlds. The new, intended option set S is the intersection of the propositional content set with the least reprehensible subnest of worlds that yields a nonempty S. The problem now is that, if, say, taking my wife is from the outset more reprehensible than taking my wine, then the meaning of

(57)

You may take my wine or (take) my wife

will effectively amount to that of (56). Provided, that is, the verb phrase is interpreted (as it should be) as the set union of predicates take my wine and take my wife. It suffices that one of the disjuncts be satisfied for the disjunction to be satisfied, and the 'minimal departure' constraint ensures this is the less reprehensible one. (In conditionals, 'minimal departure' is interpreted for a plausibility ordering.)

----------- KAMP's appeal to Grice's category of "Modus".

----

Kamp appeals to a conversational maxim, essentially Grice's (1975) 'Be brief!', to obtain both options for the CO reading.


The reasoning will be something like 'If he would have intended to permit just the wine, he could have saved himself the trouble of uttering the other disjunct'. This has a bona fide pragmatic inference overriding a pragma-semantic or semantic constraint. Moreover, the maxim easily generalizes to 'Suit form to meaning'. As indeed it would have to for the analogous cases involving an (apple) or 'free-choice' any.

Neither expression is longer than the or the smallest. The maxim thus looks set to predict just about anything. However, if such qualms are deemed insignificant, the gradient model with 'Brevity' makes predictions more specific than Kamp himself claims for it. Permissions, unlike commands, will in general turn out exclusive. (Intuitive hint: each disjunct singly is constrained to add only worlds in which the other is not realized.)

A desirable asymmetry that might sweeten the deal. But Kamp does not consider and.

In his uniform Boolean account it would denote set intersection (conjunction). The awrf-coordinated propo- 692 A. M er in Imperatives

7. Further implications and conclusions

I take for granted that any useful account of imperative meaning, no less than one for indicatives, will have to predict intuitions about coordinating connectives and quantifying determiners.

By current standards of best practice in linguistic semantics, the liberal mode of description adopted by H makes no predictions on these matters; and many of the assertions on offer that come close to doing so cannot be substantiated. So far, a formal linguistic treatment of imperatives' compositional semantics is not available in the literature. Kamp's related work is still unique in being ready to engage the data with enough rigor to permit assessment of explanatory tradeoffs. Its findings as well as its failings also pose a serious challenge to any of the currently prominent formal semantics of natural languages. For whatever the approach -- Montagovian, situationist, data-semantic -- all share an explicit or tacit commitment to largely Boolean or intuitionist interpretation of coordinating connectives, supported by Fregean separability of 'mood operator' and 'propositional content'.

Lewis's, Stalnaker's, and Kamp's work on imperatives' close relativesin-use blazed a trail last trodden on the authors' respective return trips. If the difficulties outlined above and the protagonists' proven skills are anything to go by, this is no accident. Taking a page or two out of the history of the hard sciences, one might suspect that a fertile and, for this very good reason, deep-seated commitment is involved. The commitment, crudely stated, is to a skeleton: a semantic skeleton of logic24 wrapped in conversational maxims or similar background assumptions. The former makes for intelligible recursion of meanings that involve the narrow closed-class inventory of connectives and determiners; the latter acts as an inexhaustible store of bandaid, ready to plug the next hemorrhage of data. One should feel more reluctant to offer this gory picture as a mnemonic if problems were confined to permissions and imperatives. And if formal semantics were less ready to flag out its empirical difficulties to adjunct disciplines with vastly more liberal standards of argumentative cogency.25 Of course, without a conceivable alternative for semantic recursion, the sensible reaction must be to shelve the above-mentioned difficulties as inconsequential.

Halfway to Newfoundland there is little point to calling your leaky vessel a wreck.

But do imperatives give us reason to believe that this is where we might be heading, rather than for some Bermuda Triangle or Sargasso Sea of meaning?

Here I must paddle softly.

A casual survey is no aquarium for one's own pet whale of a hunch.26

On the other hand, it would be misleading 694 A. M er in to leave an impression that the many points of detail noted in H's book and above can sustain no more than conceptual patchwork.

Indeed, it is lack of a strong-enough empirical hypothesis, rather than commitment to one, which seems to lead H into descriptive impasse. He does make observations that subvert his taxonomy and which, one suspects, might have swayed him to abandon it given more time. Thus, advice and suggestion are classed as 'accountable'; but H also notes that rejection of advice may need to be reasoned for, while rejection of a suggestion can always be willful (1987: 236); in Baier's words, principalrelated rather than principle-related. This takes a step toward recognition of the distinction between interest polarity and incentive conventions and sees H well ahead of more common-sensical speech-act taxonomies. Yet it leads him no further than to conclude, in a gesture raised into high fashion by the later Wittgenstein, that reliable distinctions are impossible to make, rather than to a reconsideration of taxonomic choices. By way of a summary let me therefore propose four aspects of imperatives that an account of their semantics could make use of: i. Imperatives exhibit nonspecificity regarding speech-act types routinely indicated by choice of modal verb in directively used indicatives. ii.

As in complements of modal main-clause indicatives, the satisfaction-conditional interpretation of connectives, and and or, appears to vary with imputed speech-act type; in view of (i), even without necessarily being conditioned by syntactic or lexical features. iii.

The speech-act typology of imperatives, as far as it is pertinent to theoretical linguistics, is structured by at least two parameters fundamental to extant formal theories of social conduct: interest (preference) and power (the ability to see one's preferences realized against opposing preferences). iv. Imperatives lack interpretive ambiguities that characterize modalcomplement indicatives conventionally used to convey directives. Points (i) and (ii) appear to inject indexicality into the last vestige of nonindexical meaning: truth conditions associated with 'logical' connectives. Its nature is rather less obvious than the tacit indexicality of addressee and time of utterance. If taxonomic salience is anything to go by, fairly complex social relations are involved (iii).

--------------- GRICE

Moreover, imperatives' lack of ambiguity along the prescriptive/descriptive divide rules out straightforward appeal to Gricean implicature by way of would-be assertoric equivalents.

Yet there are grounds for hope.

The dimensions of (iii) afford, in principle, enough relational structure for a well-established and nontrivial theory of decision and social conduct to thrive on. Why should not theoretical linguistics take a look at theoretical pragmatics so defined? Imperatives Consider, then, possible aims and means of paradigmatic speakers. Current pragmatics favors the optimally efficient exchange of information as the ostensible aim; and it takes representations based on logic to be the means. The paradigm informs the design of current theories of meaning and, by that route, the hypothesized pragma-semantic structure of languages. But it is not the only conceivable one. Classical rhetoric would look to persuasion as the paradigmatic aim, and to incentives as the means, ranging from physical threats and offers of love or money to presentation of evidence or proof. Now for the proof-theorist, particularly of an intuitionist persuasion, there are two attractive ways to consider logic. Crudely put: (i) as the residue that remains on abstracting from the contents of a successful mathematical construction (see van Dalen 1986); (ii) as specifying the minimal obligations to supply of evidence for a complex assertoric claim on someone's credulity (see Lorenzen 1962: section 2; Felscher 1986).

Both views come together if construction means proof. A proof is also a very special case of a means of persuasion. But only where it makes unmetaphoric sense to speak of proof do we have that coincidence of claim and warrant (more carefully: a function from demonstrations to theses) which justifies the constructivist move from constraints on argument to constraints on ontology. It is not at all clear that such conditions are satisfied by imperatives in general, for which the relation between possible warrants and possible claims is not a function. Gold will get almost anyone to do almost anything, but it will not establish the Schwarz inequality; nor will a proof of Zorn's lemma. If so, to presume a natural Boolean or intuitionist model theory for imperatives is more clearly (though perhaps not more of) an act of faith than doing so for indicatives. Interest polarity is crucial to the rhetorical view: there is no need to persuade someone who is eager to adopt your views, or even truly indifferent. And it supplies an immediate rationale for blatant asymmetries of claims and concessions. You will not mind getting more than you bargained for, provided it is more -- in a utilitarian sense -- of what you/bargained for, or should like to bargain for. Dually, you do not mind giving less than what you had to bargain away, provided you can get away with doing so. Suppose you make a claim. Then all you care about is not getting less than what you bargained for; it is up to the other party to set an upper bound on what you do, in fact, get. You must rationally expect it and may entertain a tacit probability distribution over where it might fall, but it is not your responsibility. And to the extent that the actual or virtual other party's upper boundary on concessions is not only compatible with, but also approaches, your claim's lower bound there is no slack. 696 A. Merin

In economic terms, we have an instance of 'pareto-optimality'. In the metapragmatics of communication this amounts to being specific or, in a well-explicated sense, informative. Purely as a side effect of what one might describe as a rational bargain. It is not difficult to see that the assumption of such a computational paradigm underlying linguistic communication -- a psychologistic or, better, anthropologistic assumption -- predicts phenomena widely handled under the label 'assertion and scalar implicature'. The aletheticassertoric case becomes a special case of a much more general scheme, and there is no need to appeal to notions of 'informativeness' which, in work on implicature, have not been explicated other than ordinally for the special case of entailment.27 But it has long been known (see O'Hair 1969) that apparent implicature arises even in those alethetic cases where no apparent entailment relations obtain; for example, as at Grice's (1961) point of departure. (61) It is red (62) It looks red to me, where (62) may implicate not-[known/accepted]-(61). So there is a case for liberating 'strength' from the genteel, nonpartisan interpretation it has acquired. The alternative interpretation, following the widely ignored and underrated work of Anscombre and Ducrot (1983), is that (61) will usually be a stronger argument for an ulterior conclusion ostensibly pursued by the speaker.


In the terms of Keynes (1922) and Carnap (1950), it will be more (positively) RELEVANT to the conclusion.


It is the tacit conclusion, as part of the context, that determines the scale and assigns ordinal rank or, indeed, a cardinal relevance value to propositions. If, as will often be the case, a certain proposition is more positively relevant than another one it entails, we again have a special case. In all cases, virtual interest polarity reflects in a duality of maximization vs. minimization of intersubjective bearers of value; that is., in dual scales. However, it remains to note that virtual interest polarity, though perhaps a better candidate than entailment-based notions for explaining implicature and 'polarity' phenomena,28 will not suffice to undo Boolean mispredictions for awrf-coordinate permissions. At least not without augmenting it to the kind of omnipotence that makes Gricean devices so unattractive to the sympathetic reader of Kant's second and third critique, and Boolean interpretation a mere obstacle to their workings. What seems wanted is a means of relating ostensible denotation and use conditions more tightly than the currently favored symbiosis of hard semantics and soft pragmatics. Imperatives There is a conceivable way out. It would be improper in this article to try to show that it is conceivable in a sense stronger than the round square or golden mountain. But it would be dereliction of duty not to say anything at all.

Perhaps a paragraph in the subjunctive mode will be ethical, addressed to readers with a mathematical background and of a pragmatically Kantian disposition toward metaphysics. Suppose we induced recursion not by lifting algebraic structure pointwise from the two-element space (2) of truth values to higher-function spaces (functions from worlds to 2, functions from entities to functions from worlds to 2, etc.). But rather by lifting from a space of rational (Q) or real (R) use values (utility values or even relevance values) to spaces of functions from contexts of use to Q or R. This would give us an ontology of linear algebras (over fields of characteristic 0) rather than Boolean or Heyting algebras. Such ontologies are familiar in decision theory, where substantive 'prospects' carrying utility are linear combinations of 'pure' (atomic) prospects that inhabit subsets -- frequently of convex cones -- of linear spaces ordered by preference. Moreover, the structure of vector spaces and their isomorphic translation spaces would make it natural to identify sentence denotata with changes to contexts, conceived of as deontic or epistemic sets of joint commitments. The hypothesis would then be that uniformly multilinear compositionality based on such ontologies informs native-speaker intuitions on meaning.29 Well before conventions of sound argument from evidence that are canonized in one's favorite logic impose or remove various constraints on inference. If serious at all, this would be an empirical hypothesis. No considerations other than those standard in the aesthetics of hard, empirical, numerate sciences should then influence its assessment. Looking at some of the more surprising lacunae in H's book and in the more formal semantic literature at large, one may indeed come to suspect the diversity of commitments that linguistics-cum-philosophy takes over from its constituent disciplines (see also Thomason 1984). Should it opt, like other empirical sciences, for elegant and comprehensive prediction of core data; here on-line intuitions? Or should it define semantics in such a way that any possible semantic theory would have to explicate more considered, reflective intuitions on sound inference? (Rather than be forced to explain how one arrives at or articulates with them by routes other than simple unwrapping of pragmatic bandages.) Or should it even tell us how languages and minds make contact or stay in touch with the Real World? That these objectives might be compatible cannot be ruled out a priori. Nor taken for granted. The brief excursion into science fiction was to suggest that a compositional, properly model-theoretic semantics of pre- 698 A. M er in dictive 'representations' not based on truth or warrant conditions and logic is at least conceivable; hence, that there might be an alternative to the otherwise reasonable policy of hanging on to a dodgy liaison come what may. Charles Hamblin's express conjecture that much of a better semantic analysis of imperatives will carry over into the semantics of other moods is, I believe, true. The question remains how to turn it into a research strategy. His book is a timely reminder that attempts to serve two disciplines at once may still serve both of them well, though not always in the way intended. It makes a strong case for amicable divorce; which is no bar to the occasional cinq a sept.


Indispensable, and reviewed in Zwicky (1987). A useful contemporary is Quirk et al. (1985: chapter 11).

An earlier useful source in transformationalist terminology is Schachter (1973).

For a government-binding view on syntax see Beukema and Coopmans (1989).

Outside mature disciplines, citation frequency is no reliable indicator of scientific pertinence.

My criterion for mandatory inclusion in a book will thus be something like, addressing roughly the same problem with a degree of formal rigor or predictive specificity not appreciably lower than that of the book itself.

That said, the list of references is still a mine of useful information.

Compare also Husserl's objection to the assertoric version of what later became the 'performative hypothesis*.

He remarks conclusively that it leads to an indefinite regress of semantically nonequivalent expressions

(1901: VI, section 68).

For recognition of this 'necessitating' nature of command directives see Kant, Metaphysics of Conduct (1797: Introduction, IV).

Why not 'quite correctly'? See section 4.

The state of the art in rational syntactic reconstruction of similar nonmonotonic changes to epistemic commitment bases closed under consequence is in Gardenfors (1988).

For a reasoned statement of such a view, often identified with Ryle and the later Wittgenstein, see Kelsen (1911).

He sees the public IMPUTATION of commitments as basic. H usefully notes that some languages (for example, Latin, Bengali) grammaticalize (amend by Or lexicalize') a distinction between immediate and later compliance (1987: 25).

In binary feature pairs: indicatives carry < + TENSE, + AUX>, infinitives and subjunctives <-TENSE, -AUX>, and imperatives <-TENSE, +AUX> (Davies 1986: 123ff.). N.B.: Davies's +AUX stands for obligatory do support in negation and emphasis.

Inspiration is Kenny (1963).

The problem posed by bare (naked) infinitives is not Imperatives addressed. But there is a nice consequence for embedded infinitival questions (... {whether/who} to ...). Check chapter 1 of your favorite text on formal languages for absence of prohibitive constraints; and, for example, Smullyan (1968: 46f.) for a reassuringly serious precedent of sorts. Compare Kant (1786: section 2).

Indeed, Paul treats as 'imperative questions' the Me vote for Head Tax?! type expressions explored in Akmajian (1984).

See Davies (1986: 248ff.) and Huddleston (1984: 361f.) for linguistic subtleties.

Also noted by Quirk et al. My terms; the latter is, above all, a legal or sociological category; something H ignores or might even be committed to deny. Whatever the label, a clean distinction is, I think, required to make sense of cofeferentiality constraints.

Schmerling (1982) makes a related point about quantifier scope, for example, for Everybody taste some dish vs. Everybody tasted some dish, but dismisses attempts at pragmatic reduction out of hand. Her tenor is formal and evolutionary: that imperative and other nonfinite clauses are poorly elaborated.

An early and useful analysis of relations between conditions and procedures of satisfaction for Imperatives is Simon (1967).

H does mention this as a problem but sees no implications. Recall section 4 and imperatives' low potential for distinctions between addressee and agent responsibility. Here, now, we have a like unambiguity for speakers. Query: is (49) permission-granting, prereporting, or both? As do Cornides (1974) and Hansson (1990), who independently offer systems of deontic logic based on preference that do not license the intuitively undesirable inferences. Neither tackles the state-changing effects of Imperatives, nor the concomitant differences between complex commands and permissions. Menger (1939), hard to find, scores two firsts: (i) definition of optative and imperative operators by consequence without initial recourse to modal notions: wishes () : = -+, where B = 'I shall be glad'; commands C(<1>) : = ~-> A,

where

A = 'Something unpleasant will happen'; (ii) restriction of these operators to domains of properly contingent ('doubtful') propositions.

The latter introduces a modality, but in a way quite different from that of usual deontic logics. With and as intersection (conjunction) and not as complement (negation).

And even the 'declarative' reading of (49) requires recourse to fairly unconstrained pragmatics to come out right. The philosopher or engineer need not be constrained by such explanatory considerations. A workable translation lore viz. rational reconstruction might do, however inelegant it looks. As distinct from more recent commitment to type-driven syntactico-semantic composition, which appears to be informed by rules instantiating those of familiar deductive systems (see Lambek 1958; Morrill and Carpenter 1990). Empirically, the emphasis here is in the first instance on judgments of well-formedness of sentences, rather than on inferential relations of pairs of sentences. Before consulting learned counsel on the law of libel, search the literature for what should be an elementary move in applied mathematics: a comprehensive check on whether native-speaker intuitions conform to the theorem schemata of Boolean algebras, distributive lattices, or even just lattices. Complete, that is, with serious attempts -- longer than the occasional one-liner on 'redundancy' -- to explain why 700 A. M er in they do not, if and where they do not. (I do not even count the case of 'collective'denoting NPs that reflect failure of the VP to distribute over and.) For an earlier freeze-dried version see Merin (1986).

Recent sources (Levinson 1987; Horn 1989) refer to the last formally phrased treatment in the open literature (Atlas and Levinson 1981).

Check that, on the information given, its INF operator symbol denotes a function mapping propositions (or sets of sentences) to themselves (or their denotata) or their (denotatas') complements.

Which raise analogous problems; check, for example, Ladusaw (1979: 4.1) on few. We should interpret binary connectives as operators Xvw[av + w] where v, w are vector variables and , scalar parameters.

For distributing and, a= = l. For or, = l -a, with {0,1} as value of a discrete random variable not under the speaker's control; under addressee's control in imperatives. For nondistributing and, as in Sammy and Rosie (between them) carried the piano upstairs, we have ]0,1[ or [0,1], that is, convex combination, as for between A and B. This would make the apparently least indexical of words (or their morphologically unrealized translations) systematically indexical. It also generalizes to w-ary connectives (Alt ... , ,,- {and/or} AH) and quantifiers.


References

Akmajian, Adrian (1984). Sentence types and the form-function fit. Natural Language and Linguistic Theory 2, 1-23.

Aqvist, Lennart (1965). Choice-offering and alternative-presenting disjunctive commands. Analysis 25, 185-187. --(1974). A new approach to the logical theory of actions and causality. In Logical Theory and Semantic Analysis, S. Stenlund (ed.), 73-91. Dordrecht: Reidel.

Anscombre, Jean-Claude, and Ducrot, Oswald (1983). L'argumentation dans la langue. Brussels: Mardaga. Atlas, Jay D., and Levinson, Stephen C. (1981). It-clefts, informativeness and logical form: radical pragmatics (revised standard version). In Radical Pragmatics, P. Cole (ed.), 1-61. New York: Academic Press. Beukema, Frits, and Coopmans, Peter (1989). A government-binding perspective on the imperative in English. Journal of Linguistics 25, 417-436.

Carnap, Rudolf (1950). Logical Foundations of Probability. Chicago: University of Chicago Press.

Chellas, Brian (1969).

The Logical Form of Imperatives. Stanford, CA: Perry Lane. --(1971). Imperatives. Theoria 37, 114-129. Cornides, Thomas (1969). Der Widerruf von Befehlen. Studium Generale 22, 1215-1263. --(1974). Ordinale Deontik: Zusammenh nge zwischen Pr ferenztheorie, Normlogik und Rechtstheorie. Vienna: Springer. Davies, Eirlys E. (1986). The English imperative. London: Croom Helm. Dowty, David R. (1979). Word Meaning and Montague Grammar. Dordrecht: Reidel. Ehrenkranz, Jean (1973). Sometimes Or' = 'and'. Linguistic Inquiry 4, 241-242. Feischer, Walter (1986). Dialogues as a foundation for intuitionistic logic. In Handbook of Philosophical Logic, vol. 3, Dov Gabbay and Franz Guenthner (eds.), 241-372. Dordrecht: Reidel. Gardenfors, Peter (1988). Knowledge in Flux. 'Cambridge, MA: MIT Press. Gazdar, Gerald (1979). Pragmatics. London: Academic Press. Imperatives --(1981). Speech act assignment. In Elements of Discourse Understanding, A. K. Joshi, B. L. Nash-Webber, and I. Sag (eds.), 64-83. Cambridge: Cambridge University Press.

Grice, Herbert Paul (1961).

The causal theory of perception. Proceedings of the Aristotelian Society 35 (suppl.), 121-152.

1971. Probability, desirability, and mood operators.

--(1975). Logic and conversation. In Syntax and Semantics 3, P. Cole and J. L. Morgan (eds.), 41-58. New York: Academic Press.

2001. Aspects of reason. On !p v !q (reviewed by Harman, online, "Mail those letters or burn them!")

---


Hamblin, Charles L. (1959). The modal 'probably'. Mind 68, 234-240. --(1970). Fallacies. London: Methuen. --(1971). Mathematical models of dialogue. Theoria 37, 133-155. --(1973). Questions in Montague English. Foundations of Language 10, 41-53. --(1987). Imperatives, Nuel Belnap (Foreword). Oxford: Blackwell. Hansson, Sven O. (1990). Preference-based deontic logic (PDL). Journal of Philosophical Logic 19, 75-93.

Hare, Richard M. (1952). The Language of Morals. Oxford: Clarendon. --(1970). Meaning and speech acts. Philosophical Review 79, 3-24. Hausser, Roland (1980). Surface compositionality and the semantics of mood. In Speech Act Theory and Pragmatics, J. R. Searle, F. Kiefer, and M. Bierwisch (eds.), 174-193. Dordrecht: Reidel. Hofstadter, Albert, and McKinsey, J. C. C. (1939). On the logic of imperatives. Philosophy of Science 6, 446-457. Horn, Laurence R. (1989). A Natural History of Negation. Chicago: University of Chicago Press. Huddleston, Rodney (1984). Introduction to the Grammar of English. Cambridge: Cambridge University Press. Huntley, Martin (1984). The semantics of English imperatives. Linguistics and Philosophy 7, 103-133. Husserl, Edmund (1901). Logische Untersuchungen (2nd ed. 1913-1921). Tübingen: Niemeyer. Kamp, Hans (1973). Free choice permission. Proceedings of the Aristotelian Society (n.s.) 74, 57-74. --(1979). Semantics versus pragmatics. In Formal Semantics and Pragmatics for Natural Languages, F. Guenthner and S. J. Schmidt (eds.), 255-287. Dordrecht: Reidel. Kant, Immanuel (1786). Grundlegung zur Metaphysik der Sitten, 2nd ed. Riga: Hartknoch. (Reprinted 1911 in Kant's Gesammelte Schriften, vol. 4. Berlin: Königl. Preuss. Akademie.) --(1797). Metaphysik der Sitten. Königsberg: Nicolovius. (Reprinted 1907 in Kant's Gesammelte Schriften, vol. 6. Berlin: Königl. Preuss. Akademie.) Kelsen, Hans (1911). über Grenzen zwischen juristischer und soziologischer Methode. Tübingen: Mohr. --(1949). General Theory of Law and State. Cambridge, MA: Harvard University Press. Kenny, Anthony (1963). Action, Emotion and Will. London: Routledge. Keynes, J. Maynard (1922). A Treatise on Probability. London: Macmillan. Ladusaw, William (1979). Polarity, Sensitivity as Inherent Scope Relations. New York: Garland. Lambek, Joachim (1958). The mathematics of sentence structure. American Mathematical Monthly 65, 154-170. Levinson, Stephen C. (1987). Minimization and conversational inference. In The Pragmatic Perspective, J. Verschueren and M. Bertucelli-Papi (eds.), 61-129. Amsterdam: Benjamins. Lewis, David K. (1973). Counterfactuals. Cambridge, MA: Harvard University Press. 702 A. M er in --(1979). A problem about permission. In Essays in Honour ofJaakko Hintikka, E. Saarinen, R. Hilpinen, I. Niiniluoto and M. Provence-Hintikka (eds.), 163-179. Dordrecht: Reidel. Lorenzen, Paul (1962). Metamalhematik. Mannheim: Bibliographisches Institut. Lyons, John (1977). Semantics. Cambridge: Cambridge University Press. McCarthy, John, and Hayes, Patrick (1969). Some philosophical problems from the standpoint of artificial intelligence. In Machine Intelligence 4, B. Meltzer and D. Michie (eds.), 463-502. Edinburgh: Edinburgh University Press. Menger, Karl (1934). Moral, Wille und Weltgestaltung: Grundlegung zur Logik der Sitten. · Vienna: Springer. (Translated 1974 as Morality, Decisions, and Social Organization: Toward a Logic of Ethics. Dordrecht: Reidel.) --(1939). A logic of the doubtful: on optative and imperative logic. Reports of a Mathematical Colloquium, 2nd series, issue l, K. Menger (ed.), 52-64. Notre Dame: Notre Dame University Press. Merin, Arthur (1986). Abstract: Or', 'and': non-boolean utility-functional connectives. Journal of Symbolic Logic 51, 850-851. Morrill, Glyn, and Carpenter, Bob (1990). Compositionality, implicational logics, and theories of grammar. Linguistics and Philosophy 13, 383-392. Nash, John F. (1953). Two-person cooperative games. Econometrica 21, 140-152. O'Hair, S. G. (1969). Meaning and implication. Theoria 35, 38-54. Paul, Hermann (1885). Prinzipien der Sprachgeschichte (4th ed. 1909). Halle: Niemeyer. Pratt, Vaughan R. (1977). Semantical considerations on Floyd-Hoare logic. 17th IEEE Symposium on Foundations of Computer Science, 109-121. New York: Institute of Electrical and Electronics Engineers. Quirk, Randolph, Greenbaum, Sidney, Leech, Geoffrey, and Svartvik, Jan (1985). A Grammar of Contemporary English. London: Longmans. Ross, Alf (1968). Directives and Norms. London: Routledge. Schachter, Paul (1973). Imperatives. In The Major Syntactic Structures of English, R. P. Stockwell, P. Schachter, and B. Partee (eds.), 633-671. New York: Holt, Rinehart and Winston. Schmerling, Susan (1982). How imperatives are special, and how they aren't. In Papers from the Parasession on Non-Declaratives, R. Schneider, K. Tuit, and R. Chametsky (eds.), 202-218. Chicago: Chicago Linguistic Society. Simon, Herbert A. (1965). The logic of rational decision. British Journal for the Philosophy of Science 16, 169-186. --(1967). The logic of heuristic decision making. In The Logic of Decision and Action, N. Rescher (ed.), 1-20. Pittsburgh: University of Pittsburgh Press. Smullyan, Raymond M. (1968). First-Order Logic. Berlin: Springer. Stalnaker, Robert C. (n.d.). Comments on Lewis's problem about permission. Unpublished maauttxipt (cited m Lew\* V919; Kamp \9!9). Thomason, Richmond H. (1984). Review of J. Barwise. Situations and Attitudes, etc. Journal of Symbolic Logic 49; 1403-1406. van Dalen, Dirk (1986). Intuitionist logic. In Handbook of Philosophical Logic, vol. 3, Dov Gabbay and Franz Guenthner (eds.), 225-339. Dordrecht: Reidel. Vendler, Zeno (1972). Res Cogitans: An Essay in Rational Psychology. Ithaca, NY: Cornell University Press. Weber, Max (1922). Wirtschaft und Gesellschaft (5th ed. 1976). Heidelberg: Mohr. Winograd, Terry (1972). Understanding Natural Language. New York: Academic Press. Zwicky, Arnold M. (1987). Review of E. Davies, The English Imperative. Linguistics 25 (4), 805-810